Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[RLlib] MADDPG: Move into main algorithms folder and add proper unit and learning tests. #24579

Merged
merged 18 commits into from
May 24, 2022

Conversation

sven1977
Copy link
Contributor

@sven1977 sven1977 commented May 7, 2022

MADDPG: Move into main algorithms folder and transfer to training_iteration API.

  • Add better unit tests for this algo.

Why are these changes needed?

Related issue number

Checks

  • I've run scripts/format.sh to lint the changes in this PR.
  • I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
  • Testing Strategy
    • Unit tests
    • Release tests
    • This PR is not tested :(

@sven1977 sven1977 changed the title [RLlib] MADDPG: Move into main agents folder and transfer to training_iteration API. [RLlib] MADDPG: Move into main algorithms folder and add proper unit and learning tests. May 22, 2022
Copy link
Member

@avnishn avnishn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does this learning test have to be run on pull requests? It feels like it doesn't impact our critical users, and can be moved to weekly.

py_test(
name = "learning_tests_two_step_game_maddpg",
main = "tests/run_regression_tests.py",
tags = ["team:ml", "tf_only", "no_tf_eager_tracing", "learning_tests", "learning_tests_discrete"],
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could we maybe instead add this to the weekly ci -- I feel like we don't need to as widely support this algorithm until there is a customer use case.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good question: I think we should move toward:

  • CI-tests: small tasks that can be learnt in 1-2 min (CartPole, TwoStepGame, maybe Pendulum or simpler cont. actions problem)
  • nightly tests: hard tasks, requiring GPU + ~1h of completion on >1 workers

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll leave this test here for now to roughly match the other algos' coverage (everyone has CartPole/Pendulum tests in the CI, which is ok).

@sven1977 sven1977 merged commit e73c37c into ray-project:master May 24, 2022
@sven1977 sven1977 deleted the maddpg_training_itr branch June 2, 2023 20:18
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants